Airborne Object Tracking Dataset

Dataset Walkthrough

🤫 Setting up

import json
import random
import os, sys
sys.path.append(os.path.dirname(os.path.realpath(os.getcwd())))
from IPython.display import display, clear_output
from random import randrange
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"]=25,25
import numpy as np

# Because Life, Universe and Everything!
random.seed(42)

def mdprint(text):
    display({
        'text/markdown': text,
        'text/plain': text
    }, raw=True)

⏱ Loading the Dataset

# Dataset for Airborne Object Tracking Dataset
from core.dataset import Dataset


notebook_path = os.path.dirname(os.path.realpath("__file__"))

local_path = notebook_path + '/part1'
s3_path = 's3://airborne-obj-detection-challenge-training/part1/'
dataset = Dataset(local_path, s3_path)

2021-04-12 10:04:54.257 | INFO     | core.dataset:load_gt:18 - Loading ground truth...

# You can add multi-part dataset as well, using below..
local_path = notebook_path + '/part2'
s3_path = 's3://airborne-obj-detection-challenge-training/part2/'
dataset.add(local_path, s3_path)

# You can add multi-part dataset as well, using below..
local_path = notebook_path + '/part3'
s3_path = 's3://airborne-obj-detection-challenge-training/part3/'
dataset.add(local_path, s3_path)

2021-04-12 10:05:11.886 | INFO     | core.dataset:load_gt:18 - Loading ground truth...
2021-04-12 10:05:31.329 | INFO     | core.dataset:load_gt:18 - Loading ground truth...

👀 DeepDive into Flight

all_flight_ids = dataset.get_flight_ids()
lucky_flight_id = random.choice(all_flight_ids)
lucky_flight_id = '7a1f085ad3834f53ae8e19cbbfe1bd36'
lucky_flight = dataset.get_flight_by_id(lucky_flight_id)

mdprint("## 🔮Lucky draw tells us to continue with: `%s`" % lucky_flight_id)

mdprint("### Let's know our flight a bit more! 🔎")
print(lucky_flight)

mdprint("This flight has **%s frames** and total **%s airborne objects**." % (lucky_flight.num_frames, lucky_flight.num_airborne_objs))

mdprint("List of Airborne Objects: ")
for airborne_obj in lucky_flight.get_airborne_objects():
    mdprint("- %s " % airborne_obj)

🔮Lucky draw tells us to continue with: 7a1f085ad3834f53ae8e19cbbfe1bd36

Let's know our flight a bit more! 🔎

Flight#7a1f085ad3834f53ae8e19cbbfe1bd36(num_frames=1199, num_airborne_objs=14)

This flight has 1199 frames and total 14 airborne objects.

List of Airborne Objects:

• AirborneObject#Airplane5(num_frames=547, planned=False)

• AirborneObject#Bird8(num_frames=19, planned=False)

• AirborneObject#Airplane2(num_frames=1046, planned=False)

• AirborneObject#Airplane3(num_frames=352, planned=False)

• AirborneObject#Airplane4(num_frames=280, planned=False)

• AirborneObject#Helicopter1(num_frames=551, planned=True)

• AirborneObject#Bird1(num_frames=1, planned=False)

• AirborneObject#Bird3(num_frames=114, planned=False)

• AirborneObject#Bird4(num_frames=116, planned=False)

• AirborneObject#Bird6(num_frames=26, planned=False)

• AirborneObject#Bird5(num_frames=114, planned=False)

• AirborneObject#Bird2(num_frames=84, planned=False)

• AirborneObject#Bird9(num_frames=21, planned=False)

• AirborneObject#Bird7(num_frames=68, planned=False)

🛫 We know text content isn't lucrative, no?!

💾 Download whole dataset (optional)

Please note, downloading image for each frame would be slower, so when using for training downloading whole dataset is preferred.

You can download full dataset using following command in your directory:

aws s3 sync s3://airborne-obj-detection-challenge-training/part1 part1/ --no-sign-request
aws s3 sync s3://airborne-obj-detection-challenge-training/part1 part2/ --no-sign-request
aws s3 sync s3://airborne-obj-detection-challenge-training/part1 part3/ --no-sign-request

💾 Download whole flight (optional)

You can download individual flight using below command:

aws s3 sync s3://airborne-obj-detection-challenge-training/part1/Images/{{flight_id}} part1/Images/{{flight_id}} --no-sign-request

or:

lucky_flight.download()

🌤 Let's quickly check how the scenery looks like in this video

image = lucky_flight.get_frame(randrange(lucky_flight.num_frames)).image()
display(image)

🖼️Well, one picture doesn't give much idea, let's try out few frames (n=9)

You can try out whole video as well if you have dataset downloaded

%matplotlib inline
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"]=25,25
fig = plt.figure()

for i in range(1, 10):
    ax = fig.add_subplot(3, 3, i)
    ax.imshow(lucky_flight.get_frame(randrange(lucky_flight.num_frames)).image(type='cv2'))

plt.show()

📉 Presence of the objects and distance in frames (when present)

airborne_objects = lucky_flight.detected_objects
frames = lucky_flight.frames

rows = len(airborne_objects.keys())

detected = {}
distance = {}
for frame_id in lucky_flight.frames:
    f = lucky_flight.get_frame(frame_id)
    for obj in airborne_objects:
        if obj not in detected:
            detected[obj] = []
            distance[obj] = []

        if obj in f.detected_objects:
            detected[obj].append(True)
            distance[obj].append(f.detected_object_locations[obj].range_distance_m)
        else:
            detected[obj].append(False)
            if not airborne_objects[obj].planned:
                distance[obj].append(float("NaN"))
            else:
                distance[obj].append(0)

i = 0
f = plt.figure(figsize=(25, 25), dpi=80)
f, axes = plt.subplots(nrows = rows, ncols = 2, squeeze=False)

for obj in detected:
    axes[i][0].scatter(range(len(distance[obj])), distance[obj])
    axes[i][0].set_xlabel('Frame')
    axes[i][0].set_ylabel('Distance')
    axes[i][0].set_title(obj)
    
    if not airborne_objects[obj].planned:
        plt.text(0.3, 0.2, "Unplanned objects\ndon't have distance", 
                 fontsize=20, transform=axes[i][0].transAxes, color="grey")

    axes[i][1].scatter(range(len(detected[obj])), detected[obj])
    axes[i][1].set_xlabel('Frame')
    axes[i][1].set_ylabel('Present')
    axes[i][1].set_title(obj)
    i += 1

# for obj in detected:
#     dt = random.choice(airborne_objects[obj].location)
#     img = dt.frame.image_annotated()
#     axes[[i, i+1]][1].imshow(img)
display(plt.show())

<Figure size 2000x2000 with 0 Axes>

None

%matplotlib inline
import matplotlib.pyplot as plt
plt.rcParams["figure.figsize"]=25,25

import time

# Let's get one of the "planned" object in this video
airborne_objects = lucky_flight.detected_objects
obj_of_interest = None
for obj in airborne_objects:
    if airborne_objects[obj].planned:
        obj_of_interest = airborne_objects[obj]
        break

images = []
fig = plt.figure()


for i in range(1, 7, 2):
    frame = obj_of_interest.location[randrange(obj_of_interest.num_frames)].frame
    ax = fig.add_subplot(3, 2, i)
    ax.imshow(frame.image(type='cv2'))
    ax = fig.add_subplot(3, 2, i + 1)
    ax.imshow(frame.image_annotated())

plt.show()

flight_path = lucky_flight.generate_video(speed_x=5)

2021-04-12 10:07:16.824 | INFO     | core.flight:generate_video:112 - Generating video...

from IPython.display import Video

Video(flight_path, embed=True, width=786)

Overall Distribution

✈️ Airborne Objects Count Distribution

num_objects = []
for flight_id in dataset.get_flight_ids():
    num_objects.append(dataset.get_flight(flight_id).num_airborne_objs)

plt.figure(figsize=(5, 5), dpi=80)
plt.hist(num_objects, density=True, bins=30)
plt.ylabel('% of data')
plt.xlabel('#airborne_objects')
display(plt.show())


print("Avg: ", sum(num_objects) / len(num_objects))
print("Min: ", min(num_objects))
print("Max: ", max(num_objects))

None

Avg:  2.5532286212914483
Min:  0
Max:  46

planned_num_objects = []
for flight_id in dataset.get_flight_ids():
    count = 0
    flight = dataset.get_flight(flight_id)
    for obj in flight.detected_objects:
        if flight.detected_objects[obj].planned:
            count += 1
    
    planned_num_objects.append(count)
    
plt.figure(figsize=(5, 5), dpi=80)
plt.hist(planned_num_objects, density=True, bins=30)
plt.ylabel('% of data')
plt.xlabel('#airborne_planned_objects')
display(plt.show())


print("Min: ", min(planned_num_objects))
print("Max: ", max(planned_num_objects))

None

Min:  0
Max:  1

obj_distance = []

for flight_id in dataset.get_flight_ids():
    flight = dataset.get_flight(flight_id)
    for obj in flight.detected_objects:
        if flight.detected_objects[obj].planned:
            for loc in flight.detected_objects[obj].location:
                obj_distance.append(loc.range_distance_m)

plt.figure(figsize=(5, 5), dpi=80)
plt.hist(obj_distance, density=True, bins=30)
plt.ylabel('% of data')
plt.xlabel('#distance')
display(plt.show())

print("Avg: ", sum(obj_distance) / len(obj_distance))
print("Min: ", min(obj_distance))
print("Max: ", max(obj_distance))

None

Avg:  1478.4207554675852
Min:  5.905805858280435
Max:  9399.925781456264

obj_center_left = []
obj_center_top = []

for flight_id in dataset.get_flight_ids():
    flight = dataset.get_flight(flight_id)
    for obj in flight.detected_objects:
        if flight.detected_objects[obj].planned:
            for loc in flight.detected_objects[obj].location:
                left, top = loc.bb.get_center()
                obj_center_left.append(left)
                obj_center_top.append(top)

heatmap, xedges, yedges = np.histogram2d(obj_center_left, obj_center_top, bins=100)
extent = [xedges[0], xedges[-1], yedges[0], yedges[-1]]

plt.figure(figsize=(10, 10), dpi=80)
plt.clf()
plt.imshow(heatmap.T, extent=extent, origin='lower')
plt.show()

# Entry??
obj_center_left_entry = []
obj_center_top_entry = []

for flight_id in dataset.get_flight_ids():
    flight = dataset.get_flight(flight_id)
    for obj in flight.detected_objects:
        if flight.detected_objects[obj].planned:
            for loc in flight.detected_objects[obj].location:
                left, top = loc.bb.get_center()
                obj_center_left_entry.append(left)
                obj_center_top_entry.append(top)
                break

heatmap, xedges, yedges = np.histogram2d(obj_center_left_entry, obj_center_top_entry, bins=50)
extent = [xedges[0], xedges[-1], yedges[0], yedges[-1]]

plt.figure(figsize=(10, 10), dpi=80)
plt.clf()
plt.imshow(heatmap.T, extent=extent, origin='lower')
plt.show()

# Exit!
obj_center_left_entry = []
obj_center_top_entry = []

for flight_id in dataset.get_flight_ids():
    flight = dataset.get_flight(flight_id)
    for obj in flight.detected_objects:
        if flight.detected_objects[obj].planned:
            loc = flight.detected_objects[obj].location[-1]
            left, top = loc.bb.get_center()
            obj_center_left_entry.append(left)
            obj_center_top_entry.append(top)

heatmap, xedges, yedges = np.histogram2d(obj_center_left_entry, obj_center_top_entry, bins=50)
extent = [xedges[0], xedges[-1], yedges[0], yedges[-1]]

plt.figure(figsize=(10, 10), dpi=80)
plt.clf()
plt.imshow(heatmap.T, extent=extent, origin='lower')
plt.show()

fin.

We hope you had fun exploring data with us! 👋

You can fork this notebook to add more explorations in starter kit here or submit your own notebook to the challenge here.